Capturing industrial CO 2 emissions in Spain: Infrastructures, costs and break-even prices

HAL Id: hal-01892620

https://hal-ifp.archives-ouvertes.fr/hal-01892620

Submitted on 18 Oct 2018

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Capturing industrial CO 2 emissions in Spain:

Infrastructures, costs and break-even prices

Olivier Massol, Stéphane Tchung-Ming, Albert Banal-Estañol

To cite this version:

Capturing industrial CO

2

emissions in Spain:

Infrastructures, costs and break-even prices

Olivier MASSOL a,b,c,d* Stéphane TCHUNG-MING Albert BANAL-ESTAÑOL c,e

Abstract

This paper examines the conditions for the deployment of large-scale pipeline and storage infrastructure needed for the capture of CO2 in Spain by 2040. It details a modeling

framework that allows us to determine the optimal infrastructure needed to connect a geographically disaggregated set of emitting and storage clusters, along with the threshold CO2 values necessary to ensure that the considered emitters will make the necessary

investment decisions. This framework is used to assess the relevance of various policy scenarios, including (i) the perimeter of the targeted emitters for a CCS uptake, and (ii) the relevance of constructing several regional networks instead of a single grid to account for the spatial characteristics of the Spanish peninsula. We find that three networks naturally emerge in the north, center and south of Spain. Moreover, the necessary CO2 break-even

price critically depends on the presence of power stations in the capture perimeter. Policy implications of these findings concern the elaboration of relevant, pragmatic recommendations to envisage CCS deployment locally, focusing on emitters with lower substitution options toward low-carbon alternatives.

Keywords: Carbon Capture and Storage; CO2 pipeline network; Break-even price for deployment. __________________________________

a IFP Énergies Nouvelles, 1-4 av. de Bois Préau, F-92852 Rueil-Malmaison, France.

b Center for Economics and Management, IFP School, 228-232 av. Napoléon Bonaparte, F-92852 Rueil-Malmaison, France. c Department of Economics, City, University of London, Northampton Square, London EC1V 0HB, UK.

d Climate Economics Chair, Palais Brongniart, 28 place de la Bourse, F-75002 Paris, France.

1. Introduction

Carbon Capture and Storage (CCS)1 is recurrently presented as a critically needed technology in long-term energy scenarios (e.g., IEA, 2017; Knopf et al., 2013)2 because it de facto reconciles the existing dependence upon fossil fuels while making achievable the ambitious CO2 abatement targets

required for a 2°C-compatible world. However, CCS faces the reality of a slower-than-anticipated uptake. With few exceptions, to date, large-scale integrated CCS projects have not been commercially deployed, and skepticism regarding the future outlook of that technology (Banks and Boersma, 2015) is now increasing. In light of these difficulties, research examining the socio-economic barriers to the deployment of CCS and proposing adapted policy remedies is now gaining momentum.3

For policymakers, a crucial question mark remains to be addressed and provides the basic motivation for the present paper: what is the market price per ton of CO2 that would be needed to

trigger the adoption of CCS capabilities? To address it, our point of departure is the policy discussion in Herzog (2011) who called for further attention to be paid to the conditions needed for the construction of a large-scale CO2 pipeline and storage infrastructure to be decided. So far, CO2 infrastructure issues

have predominantly been examined through the application of optimization techniques to identify the cost-minimizing design of an integrated CCS infrastructure network (Bakken and von Streng Velken, 2008; Middleton and Bielicki, 2009; Kemp and Kasim, 2010; Klokk et al., 2010; Mendelevitch et al., 2010; Kuby et al., 2011; Morbee et al., 2012; Oei et al., 2014; Oei and Mendelevitch, 2016). Yet, an examination of these prior contributions suggests two possible policy-relevant extensions.

1

CCS is the integrated process of capturing CO2 at large, stationary sources (e.g., thermal power plants, industrial sites) and

storing it permanently in a suitable geological formation to prevent its release into the atmosphere.

2

For example: in its latest Energy Technology Perspectives outlook (IEA, 2017), the International Energy Agency presents a 2°C scenario for which the contribution of CCS accounts for 14% (3rd position behind energy efficiency and renewable energy sources) of the additional abatements with respect to a reference scenario.

3

First, in most of these articles, the bulk of the emissions transported and stored is supplied by the power sector which is assumed to play a central role in the analysis. However, Hirschhausen et al. (2012) and Martinez Arranz (2015) question the usefulness of the deployment of carbon capture capabilities at thermal generation plants as alternative technologies (e.g., the installation of renewable energy sources) are likely to provide more affordable mitigation options. In contrast, CCS is critically needed to decarbonize other carbon-intensive industries for which there are no other abatement technologies (e.g., cement, iron and steel). As the future prospects for CCS in the power sector are jeopardized, there is a need to examine the economics of a less ambitious CCS deployment that would overlook the power sector and concentrate solely on the other industrial sectors. At first sight, one may infer that the absence of the power sector is likely to make the emergence of CCS even more complex as there will be a smaller volume of CO2 over which to spread the large fixed costs of the pipeline

infrastructure. That said, the exact nature of that effect and its magnitude still have to be documented.

The purpose of this paper is thus to examine the conditions for the deployment of a large-scale CO2 infrastructure project aimed at transporting the CO2 emissions captured at a series of industrial

clusters to a series of storage sites where the CO2 could be injected into a saline aquifer for permanent

storage. To account for the difficulty in organizing the adoption of carbon capture capabilities in the energy sector, our analysis successively considers two scenarios that depend on whether the CO2

emissions from carbon intensive industrial facilities (iron and steel, cement, pulp and paper) are supplemented or not by the volumes of CO2 captured at thermal generation plants and oil refineries.

We consider Spain as a case study for our analysis because at least three distinct lines of arguments make it an interesting candidate. First, fossil fuels represent 85% of the Spanish primary energy supply, while the 240Mt of CO2 emitted account for 7.5% of the EU28 total. As a member of the

EU, Spain is fully committed to reaching the European objectives of reducing emissions by 40% in 2030 and 85% in 2050 with respect to the 1990 levels. Second, spatial considerations cannot be overlooked. While the North Sea oil fields are recurrently presented as a preferred destination for the CO2 captured in Europe, the cost of routing the CO2 captured in Spain to the trunkline systems

envisioned in northern Europe would be prohibitive.4 Third, a remarkable data set on emission sources and storage potentials has recently been assembled for that country under the auspices of COMET, a large project funded by the EU (Boavida et al., 2011, 2013; Kanudia et al., 2013).

The paper is organized as follows. Section 2 presents the Spanish case. Section 3 describes the determination of the least-costly CCS infrastructure and identifies three regional subsystems that could be independently deployed in Spain. Section 4 has a methodological nature and explains how the conditions for CCS adoption can be identified using cooperative game-theoretic notions. Section 5 discusses our assumptions regarding the cost of the carbon capture operations conducted at each cluster. Section 6 presents our results regarding the break-even prices for the deployment of the three regional subsystems and compares them with the values derived from traditional cost-engineering studies that do

4

not take into account the existence of strategic interactions among players connected to a common infrastructure. The last section summarizes our conclusions and highlights the policy implications of our analysis. The detailed numerical assumptions retained for our analysis are presented in Appendix A. Appendix B presents the detailed specification of the optimization model used to support our analysis.

2. Background: The Spanish sources of CO

2

and potential sinks

In this section, we first describe the situation of CCS in Spain, in terms of the spatial distribution of emission clusters and storage sites, and the techno-economic characteristics of transport and storage technologies.

2.1 The emission clusters under scrutiny

According to the European Pollutant Release and Transfer Register (E-PRTR) database,5 a total of 144 large facilities are currently emitting CO2 in mainland Spain

6

in 2014. This number is too large to follow the approach in Massol et al. (2015) that examines the plants’ individual decisions to adopt carbon capture capabilities.7

Instead, we build upon the approach retained in the EU-funded COMET project and follow Boavida et al. (2011, 2013) who grouped emitters into clusters of reasonable size (see Appendix A). Our analysis thus considers 16 distinct clusters labeled E1 to E16 (see Table 1).8 The map in Figure 1 illustrates their locations. It should be noted that, with the exception of the Madrid area, these clusters are predominantly located in the coastal regions and their hinterlands, which is consistent with the spatial distributions of the country’s population and heavy industries.

Table 1. The emission clusters [PLEASE_INSERT_TABLE_1_HERE]

5

See: http://prtr.ec.europa.eu/

6

i.e.: Ceuta, Melilla, the Balearic and Canary Islands excluded.

7

It would require an evaluation of the cost of installing an optimal CCS infrastructure for each of 2144≈ 2.23×1043 coalitions that can be formed by these 144 emitters, which is computationally out of reach.

8

Figure 1. The geography of the emission clusters and the candidate pipelines and storage sites [PLEASE_INSERT_FIGURE_1_HERE]

We consider the construction of a CCS infrastructure that is aimed at being operated during a 30-year planning horizon starting in 2040. This starting date is consistent with both the IEA’s future global outlook for CCS which posits that the technology will be commercially available on that date (IEA, 2016); and the simulation results of the TIMES model developed under the COMET project that show that a mature CCS infrastructure will need to be installed in Spain on that date (Kanudia et al., 2013).

We investigate the possible future deployment of a CCS infrastructure in Spain through two scenarios. The first one is labeled “All” and posits that all the CO2 that can be captured by the thermal

power stations, the cement factories, the oil refineries, the pulp and paper plants, and the iron and steel industries at each of the 16 emission nodes will be captured. The second is labeled “Indus_only” and considers only the emissions from the cement factories, the pulp and paper plants, and the iron and steel industries at clusters where these industries collectively emit at least 1 MtCO2/year.

Our motivations for considering the restricted scenario “Indus_only” that omits both the thermal generation plants and the oil refineries are threefold. Firstly, this scenario is consistent with the Spanish coal situation. In recent years, the government has begun to emphasize the need to organize an industrial reconversion of the mining areas, aiming for a gradual closure of the coal mines (Zafrilla, 2014).9 As the Spanish coal mining industry is gradually disappearing, one wonders whether there will be a possible decline of coal-based power generation in the country. Secondly, the relevance of carbon capture technologies in the power sector is now being questioned because, in contrast to CCS, investment costs for renewables have experienced substantial cost decreases through higher learning effects which have made them a much cheaper abatement option (Hirschhausen et al., 2012; Martinez Arranz, 2015). In contrast, avoiding CO2 emissions in the industrial sectors may be more important than

in the electricity sector, because in some industries (e.g., cement, iron and steel) low-carbon substitute

9

technologies are more difficult to develop than in the electricity sector, and avoidance costs through the potential use of CCS may also be cheaper. Lastly, our decision to omit the refining sector is justified by the very high cost of equipping these industrial sites with carbon capture capabilities. An oil refinery represents a complex collection of carbon-emitting processing units that all have to be equipped with dedicated carbon capture equipment (Leeson et al., 2017). Compared to simpler industries like iron and steel or cement where the bulk of the CO2 emissions generally come from one or two sources, the

presence of a much larger number of small emission sources is reputed to make the implementation of carbon capture more technically challenging and expensive.

The “Indus_only” scenario thus echoes these recommendations to (i) abandon the aspiration of a broader deployment of CCS encompassing both the industrial activities and the energy industries, and (ii) follow a selective deployment aimed at “picking the low hanging fruit.” This would be achieved by focusing solely on the heavy industries where carbon capture is both affordable, hardly substitutable, and the least costly mitigation option.

Our assumptions regarding the annual quantities of CO2 that can be captured under the two

scenarios are based on the simulation results of the TIMES model developed by Kanudia et al. (2013) for the COMET project10 and were constructed as follows. We examine the simulation results of that model for the year 2040 in a mitigation scenario whereby the evolution of the EU energy system is obliged to achieve the EU-2020 targets as well as a 40% reduction of the domestic greenhouse gas emissions by 2050 relative to 1990.11 These simulation results provide for each emission cluster the annual quantity of CO2 that will be emitted by the electricity, pulp and paper, cement, refining, and iron

and steel plants, respectively. Only a fraction of these emissions can be captured via CCS. In this study,

10

We are grateful to Dr. Amit Kanudia (KanORS) for having kindly shared with us the detailed results of the numerous simulations he conducted for the COMET project.

11

Oei and Mendelevitch (2016) retained a similar mitigation scenario for their analysis of the deployment of CCS. We have also considered a -80% scenario as this target provides the base working assumption used by the European Commission for decarbonizing the European economy. Yet, a detailed examination of the potential for CCS deployment obtained under the -40% and -80% scenarios in the simulation results prepared for the COMET project shows that, though the timing of the deployment differ under the two scenarios, the annual quantities of CO2 captured after 2040 are similar. As we conduct a

static analysis here for 2040 when the demands for CO2 storage are similar under these two scenarios, we consider that

we use the sector-specific capture rates mentioned in Kanudia et al. (2013): 90% in the electricity and pulp and paper sectors, 85% in the cement and refining sectors, and 65% in the iron and steel sectors.12 As we did not have access to the annual emissions trajectories after the year 2040, we assume that the emission data will remain steady over time.

The annual quantities Qij of CO2 captured by the plants in each industrial sector j in each cluster

i are detailed in Figure 2 where the color blue (respectively: red, green) is used for the thermal power plants (respectively: the oil refineries, the other industries).

Figure 2. The capture potentials Qij at each cluster under each scenario (in MtCO2/year)

[PLEASE_INSERT_FIGURE_2_HERE]

Under the optimistic scenario “All,” the total annual quantity that can be captured at these 16 clusters attains 112.7 MtCO2/year which, according to the results of the TIMES-COMET model,

represents about 60% of the nation’s annual CO2 emissions at stationary sources in 2040. The power

sector (respectively, the refining sector) accounts for 39.7% (respectively, 8.9%) of that capture potential. Hence, it is interesting to highlight: (i) the important weight of the three other industrial sectors which together account for more than half of the total capture potential (58.0 MtCO2/year), and

(ii) among them, the large size of the cement sector that represents more than a third of the total capture potential (38.1 MtCO2/year). From a spatial perspective, one can remark that the distribution of the

clusters’ capture potentials is not uniform. An average cluster would have a capture potential of 7.0 MtCO2/year but the two largest clusters (namely E2 – the cement plants and heavy industries in

Asturias – and E7 the metropolitan area of Barcelona) together account for 24% of the overall capture potential whereas the two smallest clusters (namely E4 near Burgos and E16 near Santander) only capture 2.5% of that total.13

12

Overall, one can observe that the capture rates posited in our study are consistent with the rates mentioned in a recent review of the techno-economic literature on carbon capture authored by a team of researchers and CCS experts at Imperial College (Leeson et al., 2017).

13

Under the restricted scenario “Indus_only,” only 12 clusters have a capture potential larger than our 1 MtCO2/year threshold. Accordingly, there is no need to consider the emission clusters located in

Tarragona (E8), Cartagena (E11), Algeciras (E13), and Santander (E16) in that scenario. The cumulated capture potential of the industrial sites located at the 12 remaining clusters attains 55.2 MtCO2/year,

which represents 62.4% of the annual volume of CO2 emitted by the Spanish pulp and paper, cement,

and iron and steel sectors in 2040. The average quantity of CO2 that can be captured at one of these 12

clusters is 4.9 MtCO2/year but, again, the large integrated steel mill located in Asturias (E2) and the

cement factories located near Barcelona (E7) and Madrid (E15) have a significantly larger capture potential.

2.2 Storage sites

Spain has a favorable geologic endowment in onshore saline aquifers. In a recent geoscience study, Carneiro et al. (2015) examine the techno-economic characteristics (i.e., volume, injection capacities, costs) of the underground structures that could be developed in Spain. Their results indicate that it is technically possible to accumulate up to 10.3 GtCO2 without incurring a levelized cost of

storage14 larger than €7.2 per ton of CO2 injected. Building on their analysis, the present study considers

the eight cost-effective candidate storage sites mentioned in Figure 1 and listed in Table 2. That table reveals that there are substantial variations in both the capacities and costs of the storage sites. This variability reinforces the need to account for storage costs in our infrastructure planning model.15 Indeed, a simple pairing of the CO2 sources with the closest storage site may be neither feasible nor

economically efficient.

transportation corridors that exist in northern Spain, which suggests that the incremental cost of getting them connected to a pipeline infrastructure should remain reasonable. (cf. Figure 1).

14

The levelized cost of storage is calculated as the net present value of all costs over the planning horizon (i.e., our 30-year period) divided by the present value of the quantities injected over that period. The levelized cost of storage is the constant euro storage price that would be required over the 30-year period to cover all capital expenditures, subsequent periodic operating expenses, and the payment of an acceptable return to investors. It should be noted that these cost figures are based solely on technical considerations and thus omit the possibly substantial (but hard to evaluate ex ante) cost of the measures needed to make a storage project socially acceptable by the local population. In case of strong local opposition, that cost would become prohibitive and would impose the use of another storage site.

15

Table 2. The maximum injection rates and costs of the candidate storage sites [PLEASE_INSERT_TABLE_2_HERE]

2.3 Pipelines

A dedicated pipeline infrastructure is the only economically viable transportation solution that can carry the large quantities emitted by large stationary sources of CO2. In the present analysis, we

consider a predefined list of 49 candidate pipelines (cf. Appendix A) that could be installed to connect the emission clusters nodes E1 to E16 with the candidate storage nodes S1 to S8. From that list of candidate pipelines, it is possible to build a realistic network that accounts for Spain’s mountainous geography (terrain, landforms, natural transportation corridors). As shown in Figure 1, these pipelines are located along the country’s main transportation corridors.

From a cost perspective, we assume that the total cost to transport a given flow of CO2 on a

point-to-point pipeline system is directly proportional to the length of that pipeline and that the total cost per unit of distance can be decomposed into a fixed investment cost component, a variable investment cost one that is linearly varying with the transported flow of CO2 and a unit O&M cost. Regarding the

pipeline investment cost components, our approach follows the costing methodology used in Morbee et al. (2012) and is detailed in Appendix A. For concision, we simply highlight here that for a 100km-long onshore pipeline aimed at being installed on a flat terrain, we assume an annual equivalent fixed cost of €4.6 million and an annual equivalent variable cost of €0.16 per (tCO2×100 km). As indicated in

Appendix A, a correction is applied to these figures to account for the specific nature of terrain observed along each pipeline route. The obtained cost figures are thus specific to each pipeline route. Regarding O&M cost, IEA (2005) indicates that the annual operation costs vary between €1.0 and €2.5 per (tCO2×100 km). In our analysis, we use a value of €1.5 per (tCO2×100 km).

3. Optimal deployment of CCS infrastructure in Spain

infrastructure has to be analyzed as a unique integrated national infrastructure or whether it can be decomposed into a collection of regional subsystems.

3.1 The least-costly infrastructure deployment

We have formulated an optimization problem aimed at determining the least-costly design of a CCS infrastructure capable to transport and store the CO2 captured at the Spanish clusters. This model

is formally described in Appendix B. It aims at choosing the pipelines and storage sites (among our predefined and finite list of candidate pipelines and storage sites) that minimize the total annual equivalent cost of building and operating the pipeline and storage infrastructure. More precisely, we adopt a static framework for the year 2040. In that year, the total national demand for capturing CO2

and hence the annual storage requirement is imposed by our assumption. However, emitting clusters and storage sites need to be connected in a cost-effective manner. The model therefore seeks to minimize the total infrastructure cost by identifying the following optimal decisions: (i) whether, among a finite list of possible pipeline routes (linking either an emission cluster to a storage site, an emission cluster to a transit node, two emission nodes, two transit nodes, or a transit node to a storage site), a given corridor should be open, given its incurred fixed cost of deployment, and the transported quantity on that corridor given the variable operation cost; and (ii) the annual (eventually null) volume of CO2

injected in each storage site, given an exogenous, site-dependent unit cost of storage operations. As an outcome, we obtain a 2040-based static picture of the optimal – in least-cost sense – CO2 pipeline

network that matches the demand for storage with the existing capacities and possible routes.

We have run this model on the above-mentioned input data to identify the ideal CCS infrastructure under our two capture scenarios. The results are illustrated in Figure 3.

Figure 3. CO2 pipeline and storage deployment in Spain [PLEASE_INSERT_FIGURE_3_HERE]

At first sight, one could conjecture from the figures detailed below the two maps that the total annual equivalent costs of the infrastructure per unit of CO2 transported are similar in the two scenarios

the total flow transported. Yet, that first impression is misleading because there are marked differences in the cost structure of the two CCS systems. The pipeline cost figures reveal the presence of important economies of scale: the annual volume of CO2 transported in the scenario “All” is twice as large as the

one in “Indus_only” though the pipeline cost is only 66% percent larger. This result is not surprising as the total lengths of the networks are similar, which suggests that fixed pipeline costs are spread out over more units of output under the scenario “All.” In contrast, one can observe that the total storage cost is 2.77 times larger under the scenario “All” as its extra volume of CO2 saturates the capacity of the least

costly storage sites and imposes a mobilization of the more expansive ones in S1 and S8.

From a comparison of these two maps, several findings can be highlighted. First, whatever the scenario under scrutiny, the optimization model does not recommend the construction of a fully-connected national pipeline system but rather prefers the construction of a fragmented collection of pipelines that are physically disconnected. Second, the morphology of some of these pipeline connections is scenario dependent. In the north, the clusters located at Leon (E3) and Burgos (E4) either form an independent infrastructure or are embedded within a larger northern infrastructure. A similar observation can also be made for the clusters near Valencia (E10) and Almeria (12). The emissions captured in Aragon (E9) are either stored in S2 together with the ones captured along the Atlantic coast (cf., scenario “Indus_only”) or are directed to S3 where the CO2 captured in Catalunya (E7, E8) is also

directed (cf., scenario “All”). Hence, the optimal infrastructure deployment decided for the northern emission nodes and the ones located in Catalunya may not be independent. In contrast, the CO2

captured by the emitters located in the Madrid-La Mancha area (E15) is systematically routed to the neighboring storage S4 located in Cuenca, and that storage site only receives CO2 from that cluster. One

may thus wonder if a CCS deployment in E15 could be organized independently from what is decided in the other clusters.

3.2 Regional subsystems

connecting to each other, irrespectively of the choices made in other regions of the peninsula. To formally investigate this proposal and determine the boundaries of these regional subsystems, we successively consider the coalitions listed in Table 3. Each of these coalitions represents a candidate subsystem of emission areas that could potentially be separated from the national system.

Table 3. The subsystems of emission clusters under scrutiny [PLEASE_INSERT_TABLE_3_HERE]

For an emissions-transit-storage nodes subsystem, call it S, to be analyzed independently of the rest of the national system, we need to make sure that it does not interact with any other subgroup. From a cost perspective – and as described in the cost optimization model in Appendix B, this means that none of these other subgroups should be willing to join S in the course of satisfying its demand for storage because it would reduce its average cost of serving the demand. In other words, none of the costs of serving S plus another subset should be strictly subadditive, so that no economies of scale can emerge from sharing.

In formal terms, we let N denote the set of all the emission clusters considered in a given scenario and Q_i denote the total annual quantity of CO2 captured in cluster i. For each coalitions S, we evaluate

two types of costs. First, by setting Q_i =0 for the emission clusters i in the grand coalition N but not

in S (i.e., for all i∈N S\ ), we can solve the mathematical programming problem in Appendix B to evaluate C S( ) the stand-alone cost of serving S. This is the total cost of installing a pipeline and storage infrastructure optimally designed to serve the needs of the emission clusters in S. Second, we also use that optimization problem to assess the extra cost that this coalition S imposes on a coalition S' that gathers emission clusters in the remaining subset (i.e., S'⊂N S\ ). This is the incremental costs

( ') ( )'

C S∪S −C S imposed by S on S'. Accounting for all the possible non-empty coalitions S' that

can be formed by the emission clusters in N S\ and letting N\S denote the number of elements in

\

( ') ( )'

C S∪S −C S , the cost function is said to be separable because it verifies C S( ∪S')=C S( )+C S( )' .

So, if these 2N S\ −1 equality conditions hold, there are no cost interactions between the emission clusters in S and the others in N S\ and one can separately examine the deployment of a CCS infrastructure aimed at solely serving S without paying attention to the other emission clusters.

These cost comparisons16 reveal that, among all the candidate subsystems listed in Table 3, only three verify the conditions for a separable cost function: (i) all the nodes located onshore on the Atlantic coast, and the ones in Castilla-León, Aragon, and in the Mediterranean regions of Catalunya and Valencia, {E1, E2, E3, E4, E5, E6, E7, E8, E9, E10, E16}, (ii) all the nodes located in the southern regions

of Murcia and Andalucía {E11, E12, E13, E 14}, and (iii) the emission cluster located in the Madrid-La

Mancha area { }E15 . In the sequel, we thus partition the set of Spanish emission clusters into these three

subgroups and independently examine the conditions for the deployment of three autonomous CCS infrastructures that are respectively labeled: North, South and Central. An illustration of that decomposition is presented in Figure 4.

Figure 4. An illustration of the three independent subsystems [PLEASE_INSERT_FIGURE_4_HERE]

The annual volumes of CO2 captured and stored at each infrastructure and the associated

infrastructure cost under the two scenarios “All” and “Indus_only” are detailed in Table 4. The northern infrastructure has by far the largest potential for CO2 abatement. A rapid division of the total

infrastructure cost by the volume of CO2 captured and stored provides the average transportation and

storage cost. These figures indicate that the central infrastructure, which gathers a unique emission cluster, also has the lowest unit cost (less than 7€/tCO2 p.a. compared to figures larger than 9.3€/tCO2

p.a. for the other infrastructures).

Table 4. The three independent infrastructure systems [PLEASE_INSERT_TABLE_4_HERE]

16

4. Methodology: A cooperative game theoretic framework

In this section, we first provide a non-technical presentation of our cooperative game-theoretic framework. Then, two subsections detail the conditions that have to be verified for the construction of a common infrastructure to be decided. Lastly, we define the break-even CO2 price for joint CCS

adoption and show how it can be evaluated.

4.1 Cooperative game and stability notion

We consider a regional subsystem of emission clusters, like the ones identified in the preceding section, and examine the conditions for the construction of the least-cost (not necessarily fully-connected) CCS infrastructure in that subsystem, also identified in the previous section.

Hereafter, N refers to the grand coalition joining all the emission clusters in that subsystem: either

{E11, E12, E13, E14}

N = in the southern region, N ={E1, E2, E3, E4, E5, E6, E7, E8, E9, E10, E16} in the

northern region or N ={E15} in the central one.

By nature, a CO2 pipeline and storage system is a mutualized infrastructure and its cost must be

apportioned between all the individuals that feed CO2 into that system. In this paper, we assume that

each emission cluster represents an autonomous decision-making entity that can either feed all the volumes of CO2 captured by the local emitters to the grand infrastructure, feed them to a different

infrastructure or renounce CO2 capture. The arrangements guiding the internal functioning of that

emission cluster will be further discussed below. For the moment, we simply overlook that issue and treat all the emitters in a given emission cluster as a monolithic agent, that is, as an individual player.

opt for a stand-alone attitude (i.e., develop its own infrastructure). Our ambition is thus to identify whether or not it is possible to share the total cost of the grand infrastructure in such a manner that no subgroup of players has an incentive to disband. Such a cost allocation is said to belong to the “core” of the cost game.

We shall now specify what a given coalition S can achieve if it decides to opt out from the grand coalition and build an infrastructure aimed solely at serving its own needs. To do so, one should first examine the economic features of the shared elements of the CCS supply chain: the pipeline network and the storage site. Regarding transportation, the technology used in CO2 pipelines is not proprietary.

Potentially, several pipeline firms may have access to the same technology and may install a pipeline system between a group of emission clusters and some storage sites. In contrast, excludability can be at work on the storage side. At a given storage site, the quantity of CO2 that can be injected by S plus the

total volume injected by the other emission clusters in N S\ cannot exceed the capacity constraint of that storage. Such a capacity constraint creates a mutual influence among coalitions, a feature called an externality among coalitions. In the presence of externalities, the players who are about to deviate must take into account the behavior of the remaining agents because the cost incurred by the deviating coalition – thus the incentive to disband – can vary with the decisions taken by these remaining agents.

Several options can be envisaged to determine the cost incurred by a deviating coalition S that varies with the behavior posited for the remaining agents in N S\ . Most of the literature on cooperative game theory in the presence of externalities makes one of the following two extreme assumptions. Some papers assume that non-deviating members would stay together (e.g., Horn and Persson, 2001)

whereas others assume that they would split apart (e.g., Barros, 1998; Chander and Tulkens, 1997).17 In

this paper, we take the first approach because we think it is more likely that the remaining coalition members would build a joint infrastructure than they would build many, separate ones independently. The remaining members of the grand coalition are more likely to have assessed the feasibility of a joint

project (including geoscience studies, pipeline routing analyses, permitting procedures).

17

We use this observation to model the strategic behavior between the two coalitions (N S\ and S).

We assume that the remaining players in the grand coalition (i.e., the ones in N S\ ) collectively conserve

a first-mover advantage. That is, they can decide the construction of the least costly CCS infrastructure aimed at serving their own needs without taking into account the volume injected by the other emission cluster in the storage capacity constraint.

We implement these assumptions as follows. First, we determine the optimal transport and storage

decisions of the remaining coalition N S\ by solving an instance of the optimization problem in

Appendix B where the annual emissions captured by the deviating clusters are Q_i =0 for all i in S.

The solution of that mathematical programming problem provides the decision vector chosen by N S\

and thus the quantities injected at each storage site. Then, we assume that the deviating coalition S

observes the injection decisions of its complement N S\ and takes them as given by playing its best response to these injection decisions.18 By replicating that two-stage numerical procedure for each of the 2N −2 coalitions S with S⊂N and S≠ ∅ and S≠N that can be formed, we are able to determine the cost *( )

C S that would be incurred by a deviating coalition S.19 To ease the notation, we also let

( ) ( )

*

C N =C N .

4.2 The core of the cooperative cost game

We now assume that the pipeline and storage infrastructure aimed at serving the needs of the grand coalition N is supplied by a unique operator. The total cost incurred by that operator is C N( ). We let

(

1,..., N

)

r= r r where ri is the amount charged to the emission cluster i, denote the revenue vector

18

Technically, this is done by solving an adapted version of the optimization problem in Appendix B where the annual emissions captured by the remaining clusters are set to zero (i.e., Q_i =0 for all i in N S\ ) and the capacity level at each storage site is the difference between the physical capacity and the quantity already injected by the remaining players in

\

N S.

19

charged by that operator. We assume that this operator is compelled to charge a revenue vector that allows him to recover its cost and thus:

( )

* i i N r C N ∈ =

∑

. (1)

Each coalition of emission clusters S compares: _i

i S∈ r

∑

the amounts charged by the operator with

( )

*

C S the cost it would incur by deviating and adopting a stand-alone attitude. The condition for all

coalitions to rationally remain in the grand coalition is:

( )

* i i S r C S ∈ ≤

∑

, ∀ ⊂S N , S∉ ∅

{

,N

}

. (2) The set of revenue vectors that verifies conditions (1) and (2) is named the core of the cooperative cost game

(

*

)

,

N C . From an empirical perspective, it is possible to verify that the core is not empty by

using a linear programming approach similar to the one presented in Massol et al. (2015, Appendix B). The non-emptiness of the core indicates that it is possible for the infrastructure operator to charge a revenue vector that allows him to recover its cost while preventing the secession of the players.

4.3 The individual conditions required for CCS adoption

We now examine the emission clusters’ decision to adopt the proposed CCS project. We let χ_i denote the unit cost of the carbon capture operations conducted at cluster i. The definition of that unit capture cost will be further discussed in a subsequent section.

For any emission cluster i, the amount

(

)

2 CO i i

p −χ Q represents its willingness to pay for a CO2

pipeline and storage service and, thus, the amount

(

)

2

CO i i i

p −χ Q −r is its individual net benefit. Because of individual rationality, the infrastructure operator must provide a non-negative net benefit to each individual emission cluster, i.e.:

4.4 The break-even price for joint CCS adoption

The analysis in Massol et al. (2015) shows that the condition for the pipeline operator to be able to build the grand infrastructure amounts to set a revenue vector that verifies conditions (1), (2), and (3). The prevailing carbon price has a direct influence on the individual net benefit of the emission clusters and thus on the possibility for the infrastructure operator to determine such an incentive-compatible revenue vector. We thus define

2

*

CO

p the break-even price for joint CCS adoption as the critical value in the charge for CO2 emissions that would be compatible with the satisfaction of the three conditions.

This break-even price is the solution of the following linear program LP1: LP1: 2 , Min CO r p CO2 p (4) s.t. *

( )

i i N r C N ∈ =

∑

, (5)

( )

* i i S r C S ∈ ≤

∑

, ∀ ⊂S N\{∅,N}, (6)

(

pCO2 −χi

)

Qi− ≥ri 0, ∀ ∈i N. (7)

5. Data: The cost of the carbon capture operations

5.1 Data: The unit capture costs of the industrial sectors

The cost to build and operate carbon capture equipment is specific to each industrial sector (Leeson et al., 2017). In this paper, we assume the unit capture costs listed in Table 5 that are based on recent cost engineering analyses. These figures confirm that CO2 capture is extremely expensive in the

oil refining sector. In the other industrial sectors, the magnitude of the capture cost is commensurate with the ones observed in the power sector.

Table 5. The sector-specific capture costs

5.2 Cluster agreements and average vs. marginal costs

We consider two extreme assumptions regarding the arrangement guiding the internal functioning of the emission clusters. We first assume that transfers between individual plants in each cluster are feasible. In this case, the individual plants of each cluster will consider the average costs of the plants in the cluster and the overall gains to the cluster in any given coalition, as side payments can be made from the lowest to the highest cost plants to compensate the plants with higher costs. Under that first assumption, χi the unit cost of the carbon capture operations conducted at cluster i is defined as the

volume-weighted average capture cost at that cluster:

ij ij j i ij j Q Q χ χ =

∑

∑

. (8)

where Qij is the annual quantity of CO2 captured by the plants in industrial sector j in cluster i.

As side payments can be difficult to organize, we also consider a second assumption without transfers. In this case, the highest cost plant will be key as the gains obtained by any plant will need to compensate its costs. We are de facto assuming that each individual plant has veto power in the cluster. Under that second assumption, χi the unit cost of the carbon capture operations conducted at cluster i

is thus defined as follows:

Max

i ij

j

χ = χ . (9)

Of course, the reality is somewhere between these two extremes. Still, the differences between them will help us understand how crucial side payments are to the deployment of CCS in Spain.

5.3 The unit capture cost at each industrial cluster

Using the unit costs listed in Table 5, in each cluster we constructed the merit order of the local carbon capture units to calculate the unit capture cost χi at each industrial cluster under the two

above: χi is either defined as the volume-weighted average capture cost at that industrial cluster, or the

unit capture cost observed at the plant that has the most expansive carbon capture technology among all the plants in the industrial cluster. These figures are detailed in Table 6.

Observe that in some clusters the difference between the two approaches retained to evaluate the capture costs can be substantial. This difference is particularly salient under the scenario “All” for all the clusters with oil refining activities (E1, E 5, E 8, E 10, E11, E 13, E14). At these clusters, the

non-oil-refining sectors will perceive large infra-marginal rents if these players refuse to organize some side payments but that strategy also conveys the risk of substantially raising the break-even price of CO2

needed for the construction of the infrastructure.

Table 6. The unit capture cost χ_i at each cluster [PLEASE_INSERT_TABLE_6_HERE]

6. Results and discussion

We now use these unit capture costs together with the transportation and storage costs evaluated with the optimization model in Appendix B to evaluate the break-even price for joint CCS adoption. For each scenario (“All”, “Indus_only”) and each candidate infrastructure (North, Central, South), we run two instances of the linear programming problem LP1: one assuming that the unit capture cost at each industrial cluster is based on the quantity-weighted average value, and one assuming that this cost equals the marginal value, as shown in equations (8) and (9). The results are presented in Table 7.

Table 7. The break-even price for CCS adoption 2

*

CO

p (€/tCO2)

[PLEASE_INSERT_TABLE_7_HERE]

is an important finding as it suggests that CCS can remain a competitive decarbonization option even if the power sector massively opts for renewable energy sources and thus abandons the carbon capture technology. Of course, this result also suggests to concentrate on the “low hanging fruits” by selecting only the industries where the installation of carbon capture technologies is affordable (i.e., by omitting the oil refining sector).

Second, as can be expected, we can see that the required price for the deployment of a CCS infrastructure can be large, especially in the north and south subsystems that gather several emission clusters. This is particularly true for the scenario in which all plants are included (“All”), and in the case of a no-transfer agreement (“marginal”).

Third, as can be expected again, one can note that under a given scenario the break-even prices are a bit lower for the quantity-weighted case, as it avoids the veto power of the highest cost plant. It is interesting to highlight that the magnitude of the difference between the marginal and quantity-weighted cases is substantially smaller under the scenario “Indus_only.” By construction, this finding is a direct consequence of the unit capture costs listed in Table 6, yet it suggests that the detailed outcomes of the internal bargaining conducted within each cluster are likely to play a less important role under the scenario “Indus_only.”

As a side remark, we note that there is no difference between the marginal and quantity-weighted price in the north subsystem in the “Indus_only” scenario. For that specific scenario, we have closely examined the two solutions of the linear program LP1. By construction, the solution of LP1 is such that at least one of the nine constraints (7) – recall that they state that the individual net benefit of each emission cluster must be non-negative – must be binding. Interestingly, in both cases, there is a unique binding constraint: the one associated with the emission cluster E1 located in A Coruña in Galicia. Under the scenario “Indus_only,” there is only one industrial sector in that cluster (an iron and steel plant – see Figure 2) which explains why the unit capture cost at that cluster equals 57.5€/tCO2 both

the infrastructure operator when the prevailing carbon price equals

2

*

CO

p . This figure is far larger than the average cost of the infrastructure: 442/41.9 ≈ 10.55€/tCO2 suggested from the figures in Table 4. A

closer examination of the solution of LP1 reveals that this figure is exactly equal to the incremental cost that E1 imposes on the other emission clusters in the north. Hence, this is the lowest amount that can be charged by the pipeline operator without creating an opportunity for the other emitters to disband. By the way, a quick look at the location of that cluster on the map presented in Figure 4 suggests that this large incremental cost is not so surprising given the relatively small size of that cluster and its remote location (relative to those of the storage site S2).

Beyond the somehow anecdotal nature of that discussion centered on the case of the cluster E1, this analysis strongly questions the validity of the simple accounting-based or cost-engineering-based studies that evaluate the average total cost of a CCS supply chain (by simply dividing the total infrastructure cost by the total quantity stored) and implicitly presume that this figure can be interpreted as the critical price of CO2 required to trigger the construction of the CCS infrastructure. For example,

Table 8 reports two values that could be retained in these simple cost-engineering studies that overlook the complex cost interactions which exist in an infrastructure that has network characteristics. These values are: (i) the simple average cost of the infrastructure (to reflect the case where internal bargaining can be conducted within the clusters), and (ii) the sum of the unit capture cost at the most expensive plant connected to the infrastructure plus the average cost of the pipeline and storage infrastructure (to reflect the case where side-payments cannot be implemented within the clusters). While the average cost and highest capture costs are representative of the break-even price for the simple, mono-node infrastructure Central, this is clearly not the case when one considers the more complex infrastructures of North and South. For these infrastructures, the difference with the break-even price

2

*

CO

Table 8. The simple cost metrics derived from accounting reasoning (€/tCO2) [PLEASE_INSERT_TABLE_8_HERE]

7. Conclusion and policy implications

The question of how to organize the construction of a large-scale CO2 pipeline and storage system

is one of the key issues that policymakers must address to support the large-scale deployment of Carbon Capture and Storage (CCS) technologies. Previous research on that issue has two limitations that together provide the motivation for the present paper, namely (i) the potential failure of a widespread adoption of CCS in power generation and oil refining sectors, as well as (ii) the need to account for the coordination of actors along the chain to ensure a viable and mutually agreed cooperation at the regional level. This paper thus adopts a spatial approach to clarify the conditions that make the construction of a common pipeline and storage infrastructure with network characteristics a rational move for a set of regional clusters of industrial emitters that could be connected to that infrastructure. It also examines whether these conditions differ or not depending on the installation of carbon capture capabilities in the energy sector (i.e., at power plants, at oil refineries).

Taking Spain as a case study, the paper examines the least costly deployment of a national CCS infrastructure under these two scenarios. A closer analysis of their cost structures (i.e., on the separability of the cost function) reveals an important finding: this national infrastructure can be decomposed into three regionally distinct subsystems located in the north, center and south of Spain, meaning that under no circumstance of the scenarios under scrutiny does a pipeline link any pair of these regions. As these subsystems can be deployed independently, there is no need to concentrate the policymakers’ attention on the construction of a grand infrastructure with national scope, but rather a regional approach with respect to the implementation of CCS should be favored.

adoption. A comparison of these break-even prices provides a series of interesting findings from a policymaking perspective. Firstly, the non-adoption of CCS technologies in the energy sector does not make the cost of CCS prohibitive. Accordingly, the current lack of progress of CCS in the energy sector should not discourage its implementation in the other industrial sector (provided sufficient incentives can be set). Secondly, we found that the internal bargaining conducted within each cluster regarding the sharing of the carbon capture cost plays a less important role when the infrastructure stores solely the CO2 captured at industrial sites (i.e., when the energy sector is not present). We believe that this finding

results from a greater homogeneity of the sector-specific costs to implement carbon capture capabilities. Lastly, this analysis calls for further attention to be paid to the network characteristics of the CCS supply chain when trying to infer the break-even price of these infrastructures. Indeed, preliminary cost-engineering studies based on average cost concepts may substantially underestimate the true break-even price.

As in any modeling effort, we made simplifying assumptions. We, for instance, neglected the role of uncertainty regarding CO2 prices. As investments in carbon capture capabilities are irreversible, the

presence of uncertainty can influence the emitters’ individual decisions and thus the feasibility of a shared infrastructure. Risk-averse owners may thus require a higher premium to compensate for the risk of the investment. Further research could explore whether individual decisions based on a real-options framework can be combined with the cooperative game theoretic approach presented in this manuscript to gain further insights into the development of a CCS infrastructure. Incorporating the system effects of individual decisions (and thus the interactions) within a real-option framework, though, can be a challenging task.

Acknowledgements

Energy Sector” organized in Barcelona on February 7, 2017 for useful discussions. The views expressed herein are strictly those of the authors and are not to be construed as representing those of IFP Énergies Nouvelles.

References

Bakken, B.H., von Streng Velken, I., 2008. Linear Models for Optimization of Infrastructure for CO2 Capture and Storage. IEEE Transactions on Energy Conversion. 23(3), 824–833.

Banal-Estañol, A., Macho-Stadler, I., Seldeslachts, J., 2008. Endogenous mergers and endogenous efficiency gains: The efficiency defence revisited. International Journal of Industrial Organization, 26, 69–91.

Banal-Estañol, A., Eckhause, J., Massol, O., 2016. Incentives for early adoption of carbon capture technology: Further considerations from a European perspective. Energy Policy. 90, 246–252.

Banks, J.P., Boersma, T., (2015). Fostering Low Carbon Energy. Next Generation Policy to

Commercialize CCS in the United States. Issue Brief #2. Brookings Energy Security and Climate

Initiative. The Brookings Institution, Washington D.C.

Barros, P., 1998. Endogenous Mergers and Size Asymmetry of Merger Participants. Economic

Letters, 60(1), 113–119.

Boavida D., Carneiro J.F., Ramirez A., Martinez R., Czernichowski-Lauriol I., Tosato G., Rimi A., Zarhloule Y., Simoes S., Cabal H., 2011. Integrated infrastructure for CO2 transport and storage in the

west Mediterranean. GHGT-11, Amsterdam, 2440–2447.

Boavida D, Carneiro J, Martinez R, van den Broek M, Ramirez A, Rimi A, Tosato G, Gastine M., 2013. Planning CCS development in the west Mediterranean. Energy Procedia. 37:3212–3220.

Carneiro, J., Martinez, R., Suaréz, I., Zarhloule, Y., Rimi, A., 2015. Injection rates and cost estimates for CO2 storage in the west Mediterranean region. Environmental Earth Sciences. 73(6), 2951–2962.

Chander, P., Tulkens, H., 1997. The Core of an Economy with Multilateral Environmental Externalities. International Journal of Game Theory. 26, 379–401.

Comello, S., Reichelstein, S., 2014. Incentives for early adoption of carbon capture technology.

Energy Policy. 74, 579-588.

DNV, 2010. Refineries, in Global Technology Roadmap for CCS In Industry. 2010, Det Norske Veritas.

Eckhause, J.M., 2011. Methods for Employing Real Options Models to Mitigate Risk In R&D

Funding Decisions. PhD Dissertation. College Park (MD):University of Maryland.

Eckhause, J.M., Herold, J., 2014. Using real options to determine optimal funding strategies for CO2 capture, transport and storage projects in the European Union. Energy Policy. 66, 115–134.

Gough, C., O׳Keefe, L., Mander, S., 2014. Public perceptions of CO2 transportation in pipelines,

Energy Policy. 70, 106–114.

Herzog, H.J., 2011. Scaling up carbon dioxide capture and storage: From megatons to gigatons.

Energy Economics. 33(4), 597–604.

Hirschhausen, C. von, J. Herold, and P.-Y. Oei, 2012. How a “Low Carbon” Innovation Can Fail—Tales from a “Lost Decade” for Carbon Capture, Transport, and Sequestration (CCTS).

Economics of Energy & Environmental Policy. 2, 115–123.

Horn, H., Persson, L., 2001. Endogenous Mergers in Concentrated Markets. International Journal

of Industrial Organization, 19, 1213–1244.

IEA GHG, 2002. Pipeline transmission of CO2 and energy. Report Number PH 4/6. International

IEA., 2005. In Building the cost curves for CO2 storage: European sector. Technical report,

International Energy Agency, Paris.

IEA.,2016. World Energy Outlook, International Energy Agency, Paris

IEA.,2017. Energy Technology Perspectives 2017, International Energy Agency, Paris

Kanudia, A., Berghout, N., Boavida, D., van den Broek, M., Cabal, H., et al., 2013. CCS Infrastructure Development Scenarios for the Integrated Iberian Peninsula and Morocco Energy System. Energy Procedia. 37, 2645–2656.

Kemp, A.G., Kasim, A.S., 2010. A Futuristic Least-cost Optimisation Model of CO2

Transportation and Storage in the UK/UK Continental Shelf. Energy Policy. 38(7), 3652–3667.

Klokk, Ø., Schreiner, P.F., Pagès-Bernaus, A., Tomasgard, A., 2010. Optimizing a CO2 value

chain for the Norwegian Continental Shelf. Energy Policy. 38(11), 6604–6614.

Knopf, B., Chen, Y. H. H., De Cian, E., Förster, H., Kanudia, A., Karkatsouli, I., Van Vuuren, D. P., 2013. Beyond 2020—Strategies and costs for transforming the European energy system. Climate

Change Economics, 4(supp01), 1340001.

Kuby, M.J., Bielicki, J.M., Middleton, R.S., 2011. Optimal Spatial Deployment of CO2 Capture

and Storage Given a Price on Carbon. International Regional Science Review. 34(3), 285–305.

Leeson, D., Mac Dowell, N., Shah, N., Petit, C., Fennell, P.S., 2017. A Techno-economic analysis and systematic review of carbon capture and storage (CCS) applied to the iron and steel, cement, oil refining and pulp and paper industries, as well as other high purity sources. International Journal of

Greenhouse Gas Control. 61, 71–84.

Martinez Arranz, A., 2015. Carbon capture and storage: Frames and blind spots. Energy Policy. 82, 249–259.

Mendelevitch, R., Herold, J., Oei, P-Y., Tissen A., 2010. CO2 highways for Europe: Modelling a

carbon capture, transport and storage infrastructure for Europe. CEPS Working Document No. 340, Centre for European Policy Studies, Brussels, Belgium.

Middleton, R.S., Bielicki, J.M., 2009. A scalable infrastructure model for carbon capture and storage: SimCCS. Energy Policy. 37(3), 1052–1060.

Morbee, J., Serpa, J., Tzimas, E., 2012. Optimised deployment of a European CO2 transport

network. International Journal of Greenhouse Gas Control. 7, 48–61.

Oei, PY., Herold, J., Mendelevitch, R., 2014. Modeling a Carbon Capture, Transport, and Storage Infrastructure for Europe. Environmental Modeling & Assessment. 19(6), 515–531.

Oei, PY., Mendelevitch, R., 2016. European Scenarios of CO2 Infrastructure Investment. Energy

Journal. 37, Special Issue 3, 171–194.

Riesch, H., Oltra, C., Lis, A., Upham, P., Pol, M., 2013. Internet-based public debate of CCS: Lessons from online focus groups in Poland and Spain. Energy Policy, 56, 693–702.

Rosal Fernández, I. del, 2000. How costly is the maintenance of the coal-mining jobs in Europe? The Spanish case 1989–1995. Energy Policy. 28(8): 537–547.

Rubin, E.S., Davison, J.E, Herzog, H.J., 2015. The cost of CO2 capture and storage. International

Journal of Greenhouse Gas Control. 40, 78–400.

Shackley, S., McLachlan, C., Gough, C., 2004. The public perception of carbon dioxide capture and storage in the UK: results from focus groups and a survey. Climate Policy, 4(4), 377–398.

Young H.P., 1985. Methods and principles of cost allocation. In. H. Young. (Ed.) Cost Allocation: Method, principles, applications. North-Holland: Amsterdam; 1985.

Appendix A – Emission clusters and candidate pipelines

The emission clusters

We build upon the results gained from the EU-funded COMET project to group emitters into clusters of reasonable size. From a database including the technical characteristics and geographical location of the Spanish stationary sources that emitted more than 0.1 Mt CO2/a during the years 2005–

10, the COMET team conducted an exhaustive clustering exercise that resulted in the identification of 56 sources of CO2 in Spain that aggregate the emissions of the neighboring industries and power plants.

They then simulated the future emission trajectories of each of these industrial sectors, using a detailed integrated bottom-up model of the Iberian energy system based on the TIMES framework (Kanudia et al., 2013). From their simulation results,20 it appears that 23 sources account for the largest share of the nation’s industrial emissions of CO2 and offer the most promising prospects for the installation of

carbon capture capabilities. As some of these sources are geographically very close, we have further regrouped them into 16 distinct clusters labeled E1 to E16. The following table clarifies the construction of our industrial clusters from the sources of CO2considered in the COMET project.

Table A.1. Correspondence between our industrial clusters and the sources of CO2 used in the COMET project

[PLEASE_INSERT_TABLE A.1_HERE]

The candidate pipelines and their costs

Definition

Each pipeline connects two of 37 nodes: the 16 emission clusters nodes E1 to E16, the eight storage nodes S1 to S8 and the 13 intersection nodes labeled R1 to R13 that are listed in Table A.2. The later nodes represent possible network intersections between at least three pipelines. There are no CO2

injection into/withdrawal from the network at these nodes.

20

Table A.2. The intersection nodes [PLEASE_INSERT_TABLE_A.2_HERE]

Table A.3. presents the candidate pipelines, their lengths and the dimensionless average terrain-correction factors τ that will be needed to evaluate the transportation costs. In our study, the average terrain correction factors were obtained by associating each kilometer of pipeline with the values indicated in IEAGHG (2002): e.g., 1.5 for mountainous terrain, 1.1 for agricultural land.

Table A.3. The candidate pipelines [PLEASE_INSERT_TABLE_A.3_HERE]

The pipeline investment cost

We follow the standard methodology retained in CO2 pipeline models and assume that the

construction cost of a point-to-point pipeline infrastructure is directly proportional to its length. We thus consider a normalized cost per unit of length and assume that this cost can be evaluated as follows.

To evaluate the total annual equivalent investment cost of a 100km-long pipeline, we use the pipeline investment cost formula detailed in Morbee et al. (2012) to obtain the total capital expenditures and convert them into an annual equivalent cost using a 7% discount rate and assuming an infrastructure lifetime of 30 years. The annual equivalent investment cost of a 100km-long pipeline that has a steady annual output of q MtCO2/year is:

(

A₀+B q₀

)

τ , where A0 =4.6045 is the fixed cost

coefficient (in million 2015 euros),21 the variable cost coefficient is B0 =0.1641 in 2015 euros per

(tCO2×100 km) and τ is the average terrain correction factor described in IEAGHG (2002) and

detailed in Table A.2.

21

Appendix B – Designing an optimal pipeline-storage infrastructure

This Appendix details the optimization problem used to evaluate the least-cost design of a given pipeline-storage infrastructure. We first present the notations before presenting the mathematical formulation of that problem.

Notation

To begin with, we define three sets to identify the nodes of the network:

•

N

=

{

1,..., ,...,

i

N

}

the set gathering the clusters where the captured emissions are injected into the network;

•

K

=

{

1,..., ,...,

k

K

}

the set gathering the storage nodes where CO2 is withdrawn from the

network to be injected in a saline aquifer;

•

R

=

{

1,..., ,...,

r

R

}

the set of the network routing nodes that are neither connected to an emission cluster nor to a storage site. These nodes typically represent an intersection between several pipeline links.

The three sets are mutually exclusive so:

N

∩ = ∅

K

,

K

∩ = ∅

R

and

N

∩ = ∅

R

. For notational convenience, we also let

Ζ = ∪ ∪

N

K

R

denote the macro-set regrouping all the nodes and

z is used as a generic notation for a given node in

Z

. We also let

P

=

{

1,..., ,...,

p

P

}

denote the set of candidate pipeline links.

We now present the exogenous parameters.

• Qi is the total quantity captured and injected into the network at cluster

i

;

• Qk is the maximum amount of CO2 that can be withdrawn from the network to be injected